Guide
AI training for employees: a three-session plan that sticks
AI training that changes how employees work is short, uses their own tasks, and teaches the company's way rather than AI in general. Three 45-minute sessions over three weeks work for most small companies: the rules and one real task, the team's shared instructions for the work people repeat, and how to check AI output before anyone relies on it. Keep an attendance record; in the EU that also covers the AI Act's AI-literacy duty.
Short answer
- Session 1: the rules, then one real task each.
- Session 2: the team's shared instructions for repeated work.
- Session 3: checking AI output: facts, numbers, tone, confidentiality.
- Homework between sessions, and a record of who attended.
Why most companies haven't trained anyone
In Microsoft and LinkedIn's 2024 survey of 31,000 people, only 39% of those who use AI at work had received AI training from their company, and only a quarter of companies planned generative AI training that year. Meanwhile 66% of leaders said they wouldn't hire someone without AI skills. So people teach themselves, each differently, on whatever account they have.
Training pays most for the least experienced. In the NBER field study, newer support agents gained 34% in issues resolved per hour with an AI assistant, against 14% on average. Training is how you get that gain on purpose.
The three sessions
| Session | 45 minutes | Homework |
|---|---|---|
| 1. Rules and a first task | 10 min: the one-page rules. 35 min: each person does one real task from their job, briefing the AI like a new colleague. | Do the same task with AI three times; time it. |
| 2. Our shared instructions | Walk through the team's shared instructions for the three most common tasks; everyone runs each once. | Use them for a week; note where they're wrong. |
| 3. Checking AI output | Spot the errors in five real AI drafts: a wrong number, an invented fact, a confidential detail, the wrong tone. | Suggest one fix to a shared instruction. |
A worked example
Illustration: a made-up business. A 22-person property management company rolls out Claude Team. The operations director runs the three sessions on Thursday mornings.
- Session 1: the lettings team drafts a reply to a tenant's repair request; the accounts team summarizes a contractor invoice dispute. Two people find they can't paste the tenant's name and phone number into a personal account; the rules say company account only.
- Session 2: the shared instructions for repair replies, landlord monthly reports and viewing follow-ups. Everyone runs each on yesterday's real case.
- Session 3: five drafts the director planted errors in. Most people miss the invented "24-hour" repair promise in one reply, which becomes the first fix to the repair-reply instructions: never state a repair time the contractor hasn't confirmed.
The attendance sheet goes in the HR folder, dated. Three months later, new hires take the same three sessions in their first month.
Prepare it in a week
- Write the one-page rules first; training without rules teaches habits you'll undo.
- Pick the three tasks most people repeat, and write shared instructions for each.
- Collect five real AI drafts and plant one realistic error in each for session 3.
- Book three 45-minute slots a week apart, in small groups by team.
- Keep the record: date, attendees, what was covered.
Next step
After training, the shared instructions are what keep people working the same way, and instructions people have to open get skipped. The next rung is turning them into skills the AI picks by itself: how to make your first Claude skill from a doc you already have. Writing the rules for session 1? The HR guide to an employee AI policy has sample wording.
Skills for this
- AI practiceAI training plan for staffPlans three 45-minute AI sessions on each team's own tasks: the rules and a first task, the shared instructions, and checking AI output, with homework and a training record.
- AI practiceCompany AI use policy draftDrafts a short, plain-language AI use policy: approved tools, data never to paste, review and disclosure rules, owners, and points flagged for legal review.
- AI practiceReview an AI draft before useReviews an AI-generated draft before use: claims to verify, possibly invented sources, tone and policy issues, gaps against the request, and a fix list.
Questions
What should AI training for employees cover?
Three things: the company's rules (which tools, what never goes in, who checks), doing a real task from their own job with AI, and checking AI output before it's used. Generic courses on how AI works rarely change what people do on Monday.
How long should employee AI training be?
Short and spread out works better than one long day: three 45-minute sessions over three weeks, each with homework on a real task, fit around work and give people time to try it in between.
Is AI training required by law?
In the EU, Article 4 of the AI Act asks companies that use AI to take measures for their staff's AI literacy, applying since 2 February 2025. No certificate is needed; an internal record of training is enough. Elsewhere it's usually a matter of policy, not law.
AI skill
Have your AI plan your team's AI training on its own tasks
The AI training plan for staff skill has your AI plan these three sessions for your teams: their own tasks, the exercises with planted errors, and the attendance record.
14 days free · No card needed
Sources
- Microsoft and LinkedIn: 2024 Work Trend Index — only 39% of AI users got AI training from their company; 25% of companies planned generative AI training that year; 66% of leaders wouldn't hire without AI skills (read 2026-11-05)
- European Commission: AI literacy, questions and answers — Article 4 applies since 2 February 2025; amended in 2026; no certificate, internal records enough (read 2026-11-05)
- Brynjolfsson, Li and Raymond: Generative AI at Work (NBER Working Paper 31161) — 34% gain for novice workers (read 2026-11-05)
- Claude Docs: Prompting best practices — brief the AI like a new employee (read 2026-11-05)